I would never reset a watchdog timer inside a timer interrupt.

If the foreground gets locked up in a poorly envisioned I/O loop, the timer interrupt will happily interrupt the looked up routine, reset the watchdog, and return the system to the locked up state.

I guess it depends upon what you are using the watchdog timer to guard against: unexpected consequences of coding or gamma ray particles passing through the chip die. Voltage spikes and irregularities are best handled by supervisor chips.

That's what makes us hate watchdog timers... the task of strategically placing resets in such a way as to avoid making matters worse. Can we ever be sure?

aka j